Neurocase, 10(6): 452–461, 2004 Copyright © Taylor & Francis Inc. 1355-4795/04/1006–452$16.00 Neurocase Disentangling the Web: Neologistic Perseverative Errors in Jargon Aphasia Melanie S. Moses1, 2, Lyndsey A. Nickels2 and Christine Sheard1 1 School of Communication Sciences and Disorders, Faculty of Health Sciences, The University of Sydney, Sydney, Australia, and Macquarie Centre for Cognitive Science, Macquarie University, Sydney, Australia 2 Abstract This article explores the relationship between the neologisms and perseverative errors produced by KVH, a man with severe neologistic jargon aphasia. Detailed examination of KVH’s level of language processing breakdown revealed mild difficulties with phonological encoding and severe difficulties accessing the lexical form of the word. Many of KVH’s neologisms contained phonemes perseverated from previous neologisms, suggesting an integral relationship between the production of neologisms and the perseveration of phonemes. Furthermore, KVH’s patterns of whole word (total) and phonological (blended) perseverations reflected his proposed underlying language processing deficits, consistent with recent literature on perseveration (e.g., Cohen and Dehaene, 1998). However, the simple binary distinction of total and blended perseveration is proposed to be somewhat limited for understanding the underlying nature of KVH’s complex neologistic errors. Possible explanations regarding the mechanisms underlying the production of KVH’s neologistic and perseverative errors also are discussed. Introduction Neologistic errors typically characterize the language of people with jargon aphasia (Buckingham, 1987; Buckingham et al., 1978, 1979; Butterworth, 1992; Schwartz et al., 1994). However, despite their prominence in aphasia, such errors are difficult to study because they are not easily or clearly defined and there is disagreement among researchers about their underlying source. The literature has provided different criteria for defining neologisms. Some authors have defined these errors as being any nonword response (e.g., Miller and Ellis, 1987), whereas others have discriminated between “target-related neologisms,” nonwords that are phonologicallyrelated to the target, and “abstruse neologisms,” nonwords unrelated to the target, (Buckingham, 1987; Buckingham and Kertesz, 1976; Butterworth, 1992; Schwartz, et al., 1994). For the purpose of this article, we use the term neologism to refer to nonword responses that are unrelated to the target. We refer to responses that are phonologically related to the target as phonological errors. There is much debate regarding the source of neologisms. People with jargon aphasia typically demonstrate poor awareness of their speech errors, arguably making them more susceptible to the production of neologisms (Ellis et al., 1983; Marshall et al., 1998). Poor self-monitoring of speech errors in jargon aphasia has been linked with deficits in auditory comprehension (e.g., Ellis et al., 1983). However, some studies have refuted that impaired auditory comprehension is the sole cause of impaired monitoring of speech errors (e.g., Nickels and Howard, 1995). Some researchers have proposed that neologisms reflect a severe distortion of a target at a phonological encoding level, resulting in a response that no longer shares phonology with the target (e.g., Kertesz and Benson, 1970; Lecours and Lhermitte, 1969). Alternatively, neologisms may be the result of phonological distortion of an error (such as a semantic error) from an earlier stage of lexical access (e.g., Howard et al., 1985; Nickels, 2001).1 An alternative account is that neologisms are produced to fill in a “lexical” gap when word selection fails (Buckingham and Kertesz, 1976; Butterworth, 1979, 1992). Butterworth (1979, 1992) proposed that the neologisms produced by his participant, KC, may have been produced by a back-up “device” that generates pseudo-words after a failure to retrieve the target at a lexical level. He proposed that KC generated pseudo-words by random assembly of previously produced phonemes, in other words, by a process of perseveration. Butterworth (1979) found that while these neologisms obeyed English phonotactic rules, they did not reflect English phoneme frequency, supporting his theory that there was no underlying lexical target. Butterworth (1979) proposed that while KC’s phonologically-related errors reflected incomplete retrieval of the phonological form, his neologistic errors reflected an initial, failed attempt to retrieve the target word at a lexical level and a subsequent compensatory default to a Correspondence to: Dr. Melanie Moses, c/o Dr. Lyndsey Nickels, Macquarie Centre for Cognitive Science (MACCS), Macquarie University, Sydney, NSW 2109, Australia. Tel: +61 2 9664 9969; Fax: +61 2 9850 6059; E-mail: mmoses@maccs.mq.edu.au DOI: 10.1080/13554790490894057 Preservative errors in jargon aphasia neologism-generating “device.” Butterworth (1992) documented that phonemic variants of a “device” neologism may be used up to five or six times for different target words, resulting in a string of phonologically similar neologistic responses. A separate “device” with the sole purpose of generating neologisms is an understandably controversial proposal (Ellis, 1985). That withstanding, these characteristic chains of phonologically-related neologisms are a well-documented phenomenon in the literature on jargon aphasia (e.g., Brown, 1972; Buckingham et al., 1978; Butterworth et al., 1981; Green, 1969) and are demonstrated in Example 1. 453 response, may be unrelated to the target—in other words they may be neologistic (in Example 3, see responses to “iron,” “mountain,” and “baby”). We, therefore, have a convergence of the phenomena reported in the literature on perseveration, and those from the literature on neologistic jargon aphasia (e.g., Butterworth, 1979). Example 3 (San Pietro and Rigrodsky, 1986, p. 12). Picture stimulus Response Example 1 (Butterworth, 1979, p. 146). Such strings of phonologically-related neologisms already have been noted to be associated with a process of perseveration (Butterworth, 1992). Indeed, speakers with neologistic jargon aphasia frequently have been observed to perseverate on (i.e., reproduce) individual sounds or syllables from previous words or responses (e.g., Buckingham, 1985; Buckingham et al., 1978, 1979; Butterworth, 1979, 1992; Cohen and Dehaene, 1998; Schwartz et al., 1994). Santo-Pietro and Rigrodsky (1982, p. 187) referred to these types of errors as blended perseverative errors, which may involve the reproduction of multiple or single phonemes from a previous response, combined with either target-related or nonperseverative erroneous information. One type of blended perseverative error involved the carry-over of part of the phonemic structure from the immediately (or very recently) preceding response (see Example 2). Santo-Pietro and Rigrodsky (1986) proposed that entire reproductions of the preceding response might be an extreme manifestation of this phonemic carry-over. Example 2 (San Pietro and Rigrodsky, 1986, p. 9). Picture stimulus Response Another type of perseverative pattern was identified whereby long strings of phonemically-related words or intermittent recurrences of particular phoneme groups are reproduced throughout the data. The responses differ from phonemic carry-over in that they do not necessarily include phonemes from the target. Hence, as shown in Example 3, these perseverative responses, although phonologically related to a prior Despite the described prominence of neologistic errors in jargon aphasic speech (e.g., Buckingham, 1985; Buckingham et al., 1978, 1979), phoneme (or blended) perseverations often have been excluded from quantitative analyses (e.g., Allison and Hurwitz, 1967; Hudson, 1969; Martin et al., 1998). Only a few studies have attempted to quantitatively analyze the blended perseverative errors produced by speakers with aphasia to the same extent as occured for total (whole word) perseverative errors (e.g., Cohen and Dehaene, 1998; Hirsh, 1998; Santo-Pietro and Rigrodsky, 1982, 1986; c.f. Martin et al., 1998). The exclusion of blended perseverations from many previous studies may be problematic as potentially important data clearly are lost. Recent research has proposed that the types of perseverative errors produced by people with aphasia (referred to as recurrent perseverative errors by Sandson and Albert, 1984) reflect their underlying language processing impairment. For example, Cohen and Dehaene (1998) proposed that the phonemic perseverations produced by one participant arose 454 M. S. Moses et al. from his difficulties processing at a phonological encoding level in conjunction with normal amounts of persistent activation from previously activated phonemes. In contrast, they proposed that the whole word perseverations produced by another participant arose from difficulties with processing at a lexical level. Hence, it is particularly important to examine all types of perseverative errors. In light of recent research on recurrent perseveration, it seems important to examine the perseverative and neologistic errors produced by people with jargon aphasia in terms of their underlying language processing impairments. In this article, we therefore investigate the underlying nature and cause of the neologistic errors produced by KVH, a man with severe neologistic jargon aphasia. We first determine KVH’s underlying level of language processing breakdown. This is followed by a detailed examination of the relationship between KVH’s production of neologisms and his strong tendency to perseverate on previous phonemes (blended perseveration). Case Study: KVH KVH was a 71-year-old-man who suffered a left basal ganglia cerebro-vascular accident (CVA) in January 2000 and presented with severe fluent jargon aphasia.2 He was classified on the Western Aphasia Battery (WAB) (Kertesz, 1982) as having conduction aphasia (AQ = 59.6), that had resolved from an initial WAB classification of Wernicke’s aphasia. His spontaneous speech was fluent and contained much perseverative, neologistic and semantic jargon, which was rarely self-corrected. He demonstrated some difficulties comprehending more complex speech but managed well at a basic conversational level. Testing commenced when KVH was four months post-onset. Coding of Errors Only the first stressed response to each target was analyzed for classification of total numbers and types of perseverative errors (consistent with Santo-Pietro and Rigrodsky, 1986 and Hirsh, 1998). Errors were coded on three levels: 1. relationship to target (see Table 1); 2. perseverative or non-perseverative; and 3. if perseverative, as either a total or blended perseveration. Responses were coded as total perseverations if a prior response was reproduced entirely. While being a total repetition of a previous response, they also may form a part of a new compound word response (e.g., pen (→) writing pen). In line with Hirsh (1998), responses were coded as blended perseverations if 50 percent of phonemes were reproduced from a prior response in approximately the same order. In line with Santo-Pietro and Rigrodsky (1982) responses were also coded as blended perseverations if the same initial phoneme was reproduced within 5 responses, the same final phoneme within 3 responses or the same main vowel across consecutive responses. In order to establish reliability of coding, responses from one trial of each research task were independently coded by two Table 1. Target error coding criteria Error Example Lexical (real word) Semantic Formal Mixed Language Testing KVH performed a series of preliminary language tests and three research tasks: picture naming, reading aloud, and repetition. Each research task contained the same 126 items presented in different pseudo-random orders. Items were selected from the Snodgrass and Vanderwart (1980) set of 260 object pictures. Item order was controlled so that items sharing initial phonemes were separated by a minimum of four intervening items to avoid false positive identification of blended perseverations. Perseverative errors were coded according to the criteria described below. Target words varied from one to five syllables (mean = 1.76; SD = 0.91) and were selected from a range of the semantic categories classified in Snodgrass and Vanderwart (1980; see Moses et al., 2004 for stimuli). KVH performed each research task twice, with at least one week between each administration to minimize learning effects and the chance of perseverating from items across tasks. Responses across both trials were combined and are reported as proportions of a total of 252 items. KVH was encouraged to attempt only one response per item. If no response had been provided after 20 seconds, the next item was presented. Description Visual Unrelated Real word that was semantically related to target. Real word that shared either the initial phoneme or at least 50% or more phonemes with target.a Real word that was both semantically and phonologically related to target. Real word of an item similar in visual form to the target. Real word that was not related to the target in any obvious way. dog → cat dog → desk pan → pen motorcycle → bicycle orange → ball dog → apple Non-lexical (nonword) Phonological Nonword that shared either the initial phoneme or at least 50% of phonemes with target. Neologistic Nonword not reaching the criterion for phonological relatedness (i.e., sharing less than 50% of phonemes with the target and with a different initial phoneme). Nonwords that are pseudo compound words. dog → deg dog → dog→ kib ostrich → four west Other Errors Don’t know Description a Indication that response was “I don’t know” or silence unknown or if item was not responded to at all. Attempts to describe as opposed finger → when you point to name item. Multiple word jacket→ responses. Following a combination of Dell et al. (1997) and Hirsh (1998) Preservative errors in jargon aphasia researchers. Coding agreement was reviewed and recoded until a minimum of 90 percent inter-rater agreement was reached. Results Level of Language Processing Breakdown KVH’s level of language processing breakdown was interpreted within a language processing framework such as that of Nickels (2000) (similar to that of Kay et al., 1992). Table 2 displays the preliminary test scores. Preliminary language test interpretation a) Phonological processing. Both auditory analysis and auditory lexical decision were impaired. Although KVH’s hearing was adequate for speech, audiological testing found some mild high frequency loss, possibly affecting his discrimination of individual sounds. KVH was able to process some phonological Table 2. Preliminary language test scores N Raw score Proportions Phonological Processing PALPA 2 (Kay et al., 1992) (real word minimal pair discrimination) PALPA 5 (auditory lexical decision) TOTAL Nonwords High frequency Low frequency High imageability Low imageability PALPA 8 (nonword repetition) 72 55 .76 160 80 40 40 40 40 30 136 59 38 39 40 37 10 .85 .74 .95 .98 1.00 .93 .33 52 37 .71 40 37 .93 40 35 .88 60 45 .75 30 30 26 19 .87 .63 Semantic Processing Pyramids and Palm Trees (Howard and Patterson, 1992) PALPA 47 (Spoken word-picture matching) PALPA 48 (Written word picture matching) PALPA 49 (auditory synonym judgements) TOTAL High imageability Low imageability Castles Irregular and Regular Word and Nonword Reading Test (Coltheart and Leahy, 1996) TOTAL Regular words Irregular words Nonwords PALPA 25 (visual lexical decision) TOTAL Real Words Nonwords High frequency Low frequency High imageability Low imageability information in non-word repetition, evident in the fact that 90 percent of all errors were phonologically related to their targets. b) Semantics. Difficulties were evident accessing semantics via both spoken and written modalities and from pictures alone, suggesting a conceptual semantic deficit. This is further supported by his impaired performance on auditory synonym judgements. c) Orthographic processing. Impaired performance was demonstrated on both visual lexical decision and written word to picture matching, suggesting breakdown at the orthographic input lexicon and access to the already impaired semantic system. Word and nonword reading aloud were profoundly impaired. While there were no effects of regularity or lexicality on accuracy (as KVH performed so close to floor) there were more phonologically-related responses to regular words (14/30) and nonwords (13/30) than there were to irregular words (8/30).3 This suggests that KVH was more successful in accessing at least partial phonological information for regular and nonwords for which he could use sublexical processing (and was not available to the same extent for irregular words). This pattern points to an impairment accessing the phonological form of the word via a lexical reading route (see below for further discussion). Research Task Interpretation. Table 3 displays the number of correct responses in each trial of each research task. In repetition and naming, there was no difference in accuracy across trials. However, in reading aloud KVH was significantly more accurate on the second trial, possibly reflecting spontaneous recovery or cumulative effects of testing (see Nickels, 2002 for a discussion). As shown in Table 3, in repetition, KVH produced fewest errors, most of which were either phonological or formal, reflecting mild (postlexical) phonological encoding difficulties (and some probable influence of his slight impairment in auditory analysis). Few errors were neologistic. KVH was significantly better at repeating real words than nonwords (Fisher Exact Test, p = .006). In contrast, picture naming and reading aloud both elicited large numbers of errors. Approximately half of the errors Table 3. Accuracy and error types across tasks (non-perseverative and perseverative combined) Repetition Reading aloud Picture naming Accuracy trial 1 (n = 126) 70 Accuracy trial 2 (n = 126) 81 McNemar’s test trial 1 vs. trial 2 p = .178 Orthographic Processing 90 5 .06 30 30 30 2 2 1 .07 .07 .03 120 60 60 30 30 30 30 99 57 42 29 28 30 27 .83 .95 .70 .97 .93 1.00 .90 455 1 9 p = .021 8 11 p = .581 Error types # Proportion of # total errors (n = 101) Proportion of # total errors (n = 242) Proportion of total errors (n = 233) Semantic Formal Unrelated Mixed Visual Phonological Neologistic Don’t know Description 2 19 15 0 0 23 9 33 0 .02 .19 .15 .00 .00 .23 .09 .33 .00 .02 .07 .15 .01 .00 .23 .52 .00 .00 .05 .05 .27 .004 .004 .03 .47 .03 .09 5 16 36 3 0 55 127 0 0 12 11 64 1 1 6 110 6 22 456 M. S. Moses et al. produced in these tasks were neologistic (see Examples 4 and 5). A large proportion of errors in reading aloud were also phonological, reflecting KVH’s phonological encoding impairment Example 4 (Reading aloud). Example 5 (Picture naming). There were no significant effects of frequency or syllable length on accuracy in any task, although there was a significant effect of imageability in picture naming (Wald = 4.818; p = .028), further implicating a semantic impairment. KVH clearly demonstrated some difficulties in both lexical access and phonological encoding. However, before we can more specifically delineate his level of language processing breakdown, we must determine the source of his neologistic errors. A result of poor self-monitoring ability? A series of analyses were conducted to measure KVH’s ability to self-monitor his errors in each task (see Table 4).4 First, we compared the proportion of errors that KVH rejected with the proportion of rejected correct responses. Second, we compared the proportion of errors and proportion of correct responses that were reattempted. Finally, we examined the number of responses initially responded to with “don’t know,” a possible indication that a potential neologism was prevented. KVH demonstrated superior self-monitoring of his errors in repetition, the most accurate task, where few neologistic errors were produced. Proportionately more errors were rejected than in picture naming or reading aloud and he was more likely to reject an error than a correct response. He also produced the largest proportion of overt “don’t know” responses in repetition. A possible explanation for these results could be the existence of a phonological model in repetition with which KVH could compare the intended and actual response (see Butterworth, 1992). Nevertheless, even in repetition, KVH was just as likely to reattempt a correct response as he was an error and was unable to successfully self-correct his error responses. KVH reattempted only 20 percent of his error responses of which only one resulted in a correct response, consistent with previous research (Papagno and Basso, 1996; Marshall et al., 1998; Miller and Ellis, 1987). In picture naming, where large numbers of neologisms were produced, KVH reattempted significantly more error than correct responses, reflecting more accurate self-monitoring than in repetition where few neologisms were produced. These findings suggest that the relationship between neologism production and self-monitoring is far from simple and that KVH’s neologisms cannot be explained in terms of poor self-monitoring alone. A result of a severe phonological encoding impairment? A severe disruption of phonological encoding may account for the source of some of KVH’s neologisms. While recognizing that the impact of a phonological encoding deficit may result in different error patterns across tasks, if this was the primary cause of KVH’s abundant neologisms, then he should have produced large numbers of neologisms in repetition, as he did in picture naming and reading aloud. The absence of syllable length effects in any of the preliminary or research tasks further refutes that phonological encoding was Preservative errors in jargon aphasia 457 Table 4. Rejections of errors vs. correct responses, reattempts following error and correct responses, and proportion of initial “don’t know” responses out of total responses and errorsa Rejections (excluding “don’t know” and descriptions) # Rejected errors Proportion of total error responses # Rejected correct responses Proportion of total correct Rejections of error V correct Repetition Reading aloud Picture naming 30 .44 (n = 68) 28 .19 (n = 151) p < .0001*** 17 .07 (n = 242) 0 .00 (n = 10) p = 1.00 37 .18 (n = 205) 0 .00 (n = 19) p = .049* Reattempts (including ‘don’t know’ and descriptions) # Errors reattempted Proportion of total errors # Correct responses reattempted Proportion of total correct Difference between # of error vs. correct reattempted?a # Errors successfully self-corrected Proportion of total errors reattempted 20 0.20 (n = 101) 16 .11 (n = 151) p = .156 1 .05 (n = 20) 37 .15 (n = 242) 0 .00 (n = 10) p = .366 0 .00 (n = 37) 169 .73 (n = 233) 5 .26 (n = 19) p = .0001*** 1 .006 (n = 169) Don’t knows (including “don’t know” and descriptions) # Don’t know responses Proportion “don’t know”/total responses (n = 252) Proportion “don’t know”/total errors 33 .13 .33 (n = 101) 0 0 0 (n = 242) 6 .02 0.03 (n = 233) a Fisher Exact Test reported. the primary source of his neologistic errors (Butterworth, 1992). A result of an underlying lexical access impairment? An alternative explanation is that KVH’s neologistic errors resulted from a more severe impairment in accessing the lexical form of the word. In picture naming, this breakdown reflects impaired activation of semantics and subsequent insufficient activation of the phonological form. In reading aloud, impairment at the level of the orthographic input lexicon similarly results in reduced activation of the phonological form for output. KVH’s phonological encoding deficits further impact on performance in all output modalities. However, in repetition, KVH also could derive phonological information from the stimulus via a sublexical source5 (see Nickels, 1992; Hillis and Caramazza, 1995; Howard and Franklin, 1988 for further discussion). According to this “summation hypothesis,” activation of phonology via a sublexical route, combined with limited activation from semantics, acts to make it more likely that the target word will be retrieved in repetition than in naming (see Nickels, 1992; Howard and Franklin, 1988; Hillis and Caramazza, 1995 for similar accounts). As indicated by his poor nonword reading, KVH was able to derive little accurate phonological information sublexically from the written word. It is therefore proposed that when KVH is attempting to read aloud or name a picture, and the target lexical representation is insufficiently activated, then phonemes from previous responses are assembled to form a neologism. This neologism fills the lexical “slot” for the missing target, consistent with the account described by Butterworth (1979, 1992) (see below for further discussion). Perseverative influence on neologisms The majority of KVH’s neologistic errors in all tasks were perseverative (repetition: 67%; reading: 83%; naming 64%). This suggests that the production of KVH’s neologisms was strongly linked to a process of perseveration, consistent with previous literature (e.g., Buckingham, 1981, 1985; Buckingham et al., 1978, 1979; Butterworth, 1979, 1992; Schwartz et al., 1994). He produced both whole word (total) and phonological (blended) perseverative errors. These error types have been proposed to reflect different underlying language processing problems (lexical and phonological impairments, respectively, Cohen and Dehaene, 1998). It is therefore predicted that in repetition, where KVH’s phonological encoding difficulties are most evident, he should produce predominantly blended perseverative errors. In contrast, in reading aloud and picture naming, where his severe lexical processing difficulties are most evident, KVH should produce relatively more total perseverative errors (although there also may be some blended perseverative errors, reflecting the fact that these tasks also require phonological encoding). Relatively more blended perseverative errors might be predicted to occur in reading aloud than in picture naming, reflecting an additional influence of sublexical orthographic processing in the former task. Contrary to these predictions, KVH produced predominantly blended perseverative errors in all tasks, despite his significant lexical access impairment. He also produced the largest proportion of total perseverative errors in repetition (38% total, 62% blended; n = 39). Moreover, although KVH produced some total perseverative errors in picture naming (34% total, 79% blended; n = 189) and reading aloud (34% total, 66% blended; n = 133), some of these were sourced to 458 M. S. Moses et al. Table 5. Types of blended perseverative errors Task Repetition (n = 24) Reading aloud (n = 149) Picture naming (n = 88) Proportion of all (n = 261) Phonologically related (n = 54) Neologistic (n = 154) Unrelated real word (n = 45) Semantically related (n = 8) .54 (13/24) .25 (6/24) .17 (4/24) .04 (1/24) .24 (36/149) .62 (93/149) .11 (16/149) .03 (4/149) .06 (5/88) .63 (55/88) .28 (25/88) .03 (3/88) .21 (54/261) .59 (154/261) .17 (45/261) .03 (8/261) previous neologistic responses (i.e., they were total repetitions of a previous neologistic response). This challenges the theory that total perseverative errors reflect impairment at a lexical level as neologisms, being nonwords, by definition have no lexical representation. Perhaps there was something specific to KVH’s perseverative errors that might explain these seemingly paradoxical results. KVH appeared to produce two main types of blended perseverative errors, those that were phonologically related to the target (i.e., formal or phonological errors) and those that were unrelated to the target (i.e., either neologistic or unrelated real word errors; see Table 5).6 Blended perseverative errors that were phonologically-related to the target KVH occasionally perseverated on a single or a small number of phonemes from a recent response, whilst preserving enough target-related phonological information to form a phonologically related error (see Example 6). KVH produced the largest proportion of these errors in repetition (see Table 5). Proportionately fewer of these errors were produced in reading aloud and they were virtually absent from picture naming. All of the phonologically related blended perseverative errors in repetition shared a minimum of 50 percent of phonemes in common with the target, suggesting a distortion of an activated lexical representation and less reliance on previously activated phonological information. Example 6 Blended perseverative errors that were unrelated to the target The most common blended perseverative errors produced were neologistic. These were particularly prominent in picture naming and reading aloud, where they formed strings of phonologically-related neologisms. While the sources of KVH’s phonologically-related perseverative errors were relatively clear, these errors did not appear to arise from any particular source. Rather, they tended to be phonologically related to multiple prior neologistic perseverative errors, as shown previously in Examples 4 and 5. These blended perseverative errors are clearly different from those produced in repetition. The errors are most consistent with the characteristic strings of phonologicallyrelated neologisms. These are proposed to reflect underlying lexical processing deficits in jargon aphasia (e.g., Buckingham, 1987; Butterworth, 1979, 1992), exactly the level of impairment we have argued for KVH in these tasks. KVH also produced an unpredictably large proportion of total perseverative errors in repetition, where his lexical impairments were argued to be less evident (see discussion earlier). This appears to challenge Cohen and Dehaene’s (1998) proposal that total perseverative errors reflect impaired lexical processing. Santo-Pietro and Rigrodsky (1986) proposed that total perseverative errors might be an extreme manifestation of phonemic carry-over. Accordingly, we would predict that total perseverative errors would be sourced to the immediately preceding responses. In fact, 93 percent (14/15) of KVH’s total perseverative errors in repetition were sourced to the immediately preceding response, suggesting that these errors may in fact have been reproduced from persisting activation at the level of phonological encoding. A further challenge to current theory regarding the source of total perseverations was that in reading aloud and picture naming KVH produced total perseverations of neologisms. Clearly, as argued by Hirsh (1998), neologistic total perseverations cannot be the result of increased competition at a lexical level as neologisms have no lexical representation. What then could be the source of these neologistic total perseverations? a) Phonological distortion of a real word. One possibility is that the original neologism could have been a phonological distortion of a word, due to phonological encoding difficulties following accurate lexical retrieval. Reproduction of the same underlying target could therefore be subject to the same degree and type of phonological distortion, resulting in the neologism being reproduced. However, as KVH produced sequences of similar variations of the same neologism for various consecutive (and most often) unrelated targets, it is highly unlikely that the same underlying lexical item had been retrieved and subsequently distorted in response to the different targets (see also Butterworth, 1992). b) Chance total reproductions. Perhaps a more plausible explanation is that an entire phonological sequence of a Preservative errors in jargon aphasia neologism was reproduced by chance in response to a subsequent item. As demonstrated in Example 4 (earlier) from reading aloud, initially is produced in response to the target “chain” and is later completely reproduced in response to the target “mountain.” However, it is clear that is also phonologically linked to the intervening responses, and . This explanation is again consistent with the phonologically similar strings of successive neologistic responses, characteristic of people with neologistic jargon aphasia. Butterworth (1979, 1992) and Butterworth et al., (1981) maintained that phonemes from previous neologisms remain active in a buffer for a short period of time and are randomly reassembled to form subsequent neologisms. Accordingly, it is plausible that KVH could randomly reproduce the entire phonological sequence of a prior neologism within a small number of intervening items. This is supported by the fact that the majority (62%) of his neologistic total perseverations occurred within five intervening items. Persisting activation at a phoneme level within other theories (e.g. Dell et al., 1997) can account for total perseverations of neologisms from immediately preceding responses. Indeed, this is the mechanism we propose to be most plausible for KVH’s phonologically-related blended perseverations and total perseverations in repetition. However, such theories are unable to account for perseveration where there are intervening responses, as the activation of the intervening response “wipes out” any persisting activation from the earlier response; hence, Butterworth’s (1979, 1992) requirement for a separate buffer (see also Hirsh, 1998, for similar a conclusion). General Discussion Neologistic errors characterise the verbal output of people with jargon aphasia; however, there remains considerable debate in the literature as to their underlying source. In addition, few studies have investigated the relationship between the production of neologisms and the occurrence of perseverations. Here we have addressed both these issues in a detailed single case study of KVH, a man with jargon aphasia. KVH’s speech production was impaired at the level of phonological encoding; however, this could not explain the large numbers of neologisms produced in picture naming and reading (in comparison to repetition). These neologisms were argued, instead, to reflect impaired activation of phonological forms. While consistent with certain studies on jargon aphasia (Butterworth, 1979, 1992; Simmons and Buckingham, 1992), this proposal contradicts those studies that have proposed that neologisms reflect severe underlying phonological encoding difficulties alone (Kertesz and Benson, 1970; Lecours and Lhermitte, 1969). We note that KVH’s neologistic errors were typical of those observed in speakers with jargon aphasia and that in many ways, KVH’s neologistic errors appear consistent with Butterworth’s (1979, 1992) neologism generator theory. However, we acknowledge the scepticism surrounding such a theory that appears at odds with cognitive neuropsychological principles by proposing a module that has a primary role 459 in error production (e.g., Ellis, 1985). An alternative account is that KVH’s neologistic perseverations reflect a substitution of phonemes based on the frequency of these phonemes in the language, given inadequate activation of any node at the phoneme level—a “default” mechanism (Butterworth, 1992). Hence, the appearance of perseveration is, in fact, due to the repeated use of the most frequent phonemes in the language. Butterworth (1979) argued against this, on the grounds that the phonemes in KC’s neologisms did not follow the phoneme frequency distribution of English. However, we should consider the possibility that as KC had jargon aphasia for some time at the point at which Butterworth studied his performance, it is possible that his personal phoneme frequency did not mirror that of the language. In other words, over time, those phonemes that were used in his neologisms (e.g., ) would become more frequent (e.g., long-term changes in resting levels of activation) and hence be more likely to be substituted.7 While beyond the scope of this investigation, this is clearly an area for further investigation. Through our examination of both total and blended perseverative errors, we have demonstrated a strong and quantifiable link between KVH’s production of neologisms and the perseveration of phonemes from prior responses. We would argue that as blended perseverative errors are so integrally linked with the production of neologisms, they can and must be included when studying the perseverative errors of speakers with jargon aphasia. Furthermore, we have demonstrated that the distributions of these different types of blended perseverative errors produced across the tasks reflected KVH’s level of language processing breakdown, supporting recent research (e.g., Cohen and Dehaene, 1998). However, we have demonstrated that it was overly simplistic to assume that a single functional lesion underlines a classification of “blended perseveration” as this term can encompass different error types. Similarly, the classification “total perseveration” may encompass errors of more than one type, with different production mechanisms. In repetition, an underlying phonological encoding impairment may result in perseveration of phonemes from a recent or immediately preceding response. This usually will result in blended perseverations but occasionally all the phonemes of a previous response may be perseverated, resulting in a total perseveration. In reading and naming, KVH’s severe impairment of lexical access results in a different kind of total perseveration, which is neologistic and often from a source other than the immediately preceding response. We conclude that when examining the source of neologisms, both the target error relationship and the relationship with previous responses (i.e., whether or not errors are perseverative) are integrally linked and to study one without the other inevitably leads to an inadequate description of the data. Moreover, both relationships need to be accounted for in any adequate account of neologistic jargon aphasia. While our findings appear consistent with other recent case studies (e.g., Hirsh, 1998), the conclusions drawn would be further strengthened by replication across a series of individuals with jargon aphasia. 460 M. S. Moses et al. Acknowledgements We express our greatest appreciation to the late KVH for his enthusiasm and many hours of hard work. We would like to thank Professor David Howard for his statistical support in this article, and Karalyn Patterson and anonymous reviewers for their helpful comments on earlier versions of this paper. During the preparation of this article, Dr. Lyndsey Nickels was funded by an Australian Research Council QE2 Fellowship. Notes 1 While some authors also have suggested that neologisms could be the result of a permanently and severely corrupted lexical representation, in some theories the distinction between phonological storage and encoding is not maintained (e.g., Dell, 1986), while in others such an impairment would result in a semantic error rather than phonological (e.g., Nickels, 2000). Indeed it is hard to conceive quite how such an impairment might be implemented computationally and, hence, we do not discuss it further in this article, but rather focus on the well-accepted phonological encoding account. 2 Aphasia may seem a surprising symptom of a basal ganglia infarct; however, nonthalamic subcortically originating aphasias have been found to present similarly to cortical aphasias (e.g., Nadeau and Gonzalez-Rothi, 2001). 3 When phonologically-related and correct responses were combined, the difference in performance between regular and irregular words approached significance (Fisher Exact Test, p = .096). 4 As only the initial response attempts were coded the accuracy of coding is not affected by subsequent rejections of errors. 5 We recognize alternative accounts where no separate lexical and sublexical routes are specified for reading aloud or repetition. Rather, there is a single source of phonological representation, and the degree to which this is activated is dependent on the mapping from different stimulus modalities and the degree of transparency (e.g., see Patterson et al., 1998). However, we argue that both accounts would predict the same net effect on KVH’s error patterns across the different language tasks. 6 As they accounted for only 3 percent (8/261) of all of KVH’s blended perseverative errors, semantically related blended perseverative errors are not discussed. 7 It is important to note that this frequency or long-term cumulative priming-based account is different from the short-term priming account of total perseverations rejected above (see discussion of Dell et al., 1997.) Most theories make the distinction between the two. References Allison RS, Hurwitz LJ. On perseveration in aphasics. Brain 1967; 90: 429–48. Brown J. Aphasia, apraxia and agnosia: Clinical and theoretical aspects. Springfield, Illinois: Thomas; 1972. Buckingham HW Jr. Phonological aspects of aphasia. Topics in Language Disorders 1981: 29–39. Buckingham HW. Perseveration in aphasia. In: Newman S, Epstein R, editors. Current perspectives in dysphasia. Edinburgh: Churchill Livingston, 1985: 113–54. Buckingham HW. Phonemic paraphasias and psycholinguistic production models for neologistic jargon. Aphasiology 1987; 1: 381–400. Buckingham HW, Jr, Kertesz A. Neologistic jargon aphasia. Amsterdam: Swets & Zeitlinger; 1976. Buckingham HW Jr, Whitaker HA, Whitaker, H.A. Alliteration and assonance in neologistic jargon aphasia. Cortex 1978; 14: 365–80. Buckingham HW Jr, Whitaker HA, Whitaker HA. On linguistic perseveration. Studies in Neurolinguistics 1979; 4: 329–35. Butterworth B. Hesitation and the production of verbal paraphasias and neologisms in jargon aphasia. Brain and Language 1979; 8: 133–61. Butterworth B. Disorders of phonological encoding. Cognition 1992; 42: 261–86. Butterworth B, Swallow J, Grimston M. Gestures and lexical processes in jargonaphasia. In: Brown JW, editor. Jargonaphasia. New York: Academic Press, 1981: 113–24. Cohen L, Dehaene S. Competition between past and present: Assessment and interpretation of verbal perseverations. Brain 1998; 121: 1641–59. Coltheart M, Leahy J. Assessment of lexical and non-lexical reading abilities in children: Some normative data. Australian Journal of Psychology 1996; 48: 136–40. Dell G. A spreading-activation theory of retrieval in sentence production. Psychological Review 1986; 93: 283–321. Dell GS, Schwartz MF, Martin N, Saffran EM, Gagnon D. Lexical access in aphasic and nonaphasic speakers, Psychological Review 1997; 104: 801–38. Ellis AW. The production of spoken words: A cognitive neuropsychological perspective. In: Ellis AW, editor. Progress in the psychology of language, vol. 2. London: Lawrence Erlbaum Associates, 1985. Ellis AW, Miller DM, Sin G. Wernicke’s aphasia and normal language processing: A case study in cognitive neuropsychology. Cognition 1983; 15: 111–44. Green E. Phonological and grammatical aspects of jargon in an aphasic patient. Language and Speech 1969; 12: 103–18. Hillis AE, Caramazza A. Converging evidence for the interaction of semantic and sublexical phonological information in accessing lexical representations for spoken output. Cognitive Neuropsychology 1995; 12: 187–227. Hirsh KW. Perseveration and activation in aphasic speech production. Cognitive Neuropsychology 1998; 15: 377–88. Howard D, Franklin S. Missing the meaning? Cambridge, MA: MIT Press; 1988. Howard D, Patterson K. Pyramids and palm trees test. Bury, St. Edmunds: Thames Valley Test; 1992. Howard D, Patterson K, Franklin S, Orchard-Lisle V, Morton J. Treatment of word retrieval deficits in aphasia. A comparison of two therapy methods. Brain 1985; 108: 817–29. Hudson AJ. Perseveration. Brain 1969; 91: 571–82. Kay J, Lesser R, Coltheart M. Psycholinguistic Assessment of Language Processing in Aphasia. Hove, UK: Lawrence Erlbaum Associates; 1992. Kertesz A. Western Aphasia Battery. NY: Grune and Stratton; 1982. Kertesz A, Benson DF. Neologistic jargon: A clinico-pathological study. Cortex 1970; 6: 362–86. Lecours AR, Lhermitte F. Phonemic paraphasias: Linguistic structures and tentative hypotheses. Cortex 1969; 5: 193–228. Marshall J, Robson J, Pring T, Chiat S. Why does monitoring fail in jargon aphasia? Comprehension, judgement and therapy evidence. Brain and Language 1998; 63: 79–107. Martin N, Roach A, Brecher A, Lowery J. Lexical retrieval mechanisms underlying whole-word perseveration errors in anomic aphasia. Aphasiology 1998; 12: 319–33. Miller D, Ellis A. Speech and writing errors in “neologistic jargonaphasia”: A lexical activation hypothesis. In: Coltheart M, Job R, Sartori G, editors. The cognitive neuropsychology of language. Hillsdale, NJ: Lawrence Erlbaum Associates, 1987; 253–70. Moses MS, Sheard C, Nickels LA. Insight into recurrent perseverative errors in aphasia: A case series approach. Manuscript submitted for publication; 2004. Nadeau SE, Gonzalez-Rothi LJ. Rehabilitation of subcortical aphasia. In: Chapey R, editor. Language intervention strategies in aphasa and related neurogenic communication disorders, 4th edition. Philadelphia, PA: Lippincott Williams and Wilkins, 2001; 457–71. Nickels L. Improving word finding: Practice makes (closer to) perfect. Aphasiology 2002; 16: 1047–60. Preservative errors in jargon aphasia Nickels L. Words fail me: Symptoms and causes of naming breakdown in aphasia. In: Berndt RS, editor. Handbook of neuropsychology, 2nd edition (vol. 3). Language and Aphasia. Amsterdam: Elsevier, 2001; p. 115–35. Nickels LA. A sketch of the cognitive processes involved in the comprehension and production of single words, 2000. Retrieved 1/6/04 from http:// www.maccs.mq.edu.au/˜lyndsey/model.doc. Nickels LA. The autocue? Self-generated phonemic cues in the treatment of a disorder of reading and naming. Cognitive Neuropsychology 1992; 9: 155–82. Nickels L, Howard D. Phonological errors in aphasic naming: Comprehension, monitoring and lexicality. Cortex 1995; 31: 209–37. Papagno C, Basso A. Perseveration in two aphasic patients. Cortex 1996; 32: 67–82. Patterson K, Okada S, Suzuki T, Ijuin M, Tatsumi I. Fragmented words: A case of late-stage progressive aphasia. Neurocase 1998; 4: 219–30. 461 Sandson J, Albert ML. Varieties of perseveration. Neuropsychologia 1984; 22: 715–32 Santo-Pietro MJ, Rigrodsky S. The effects of temporal and semantic conditions on the occurrence of the error response of perseveration in adult aphasics. Journal of Speech and Hearing Research 1982; 25: 184–92. Santo-Pietro MJ, Rigrodsky S. Patterns of oral-verbal perseveration in adult aphasia. Brain and Language 1986; 29: 1–17. Schwartz MF, Saffran, EM, Bloch DE, Dell G. Disordered speech production in aphasic and normal speakers. Brain and Language 1994; 47: 52–88. Simmons N, Buckingham HW. Recovery in jargon aphasia, Aphasiology 1992; 6: 403–14. Snodgrass JG, Vanderwart M. A standardised set of 260 pictures: Normals for name agreement, image agreement, familiarity, and visual complexity. Journal of Experimental Psychology: Human Learning and Memory 1980; 6: 174–215.